Open ljnelson opened 5 years ago
I raise this because vegeta reports that there was a timeout while reading the body:
Vegeta has a default -timeout
of 30s
. I think curl
has no default timeout. Set -timeout=0
and try again.
The requests being different is likely due to different request headers being sent (e.g. User-Agent
).
Version and Runtime
Expected Behaviour
I wanted to have
vegeta
issue a "plain text" HTTP/2 call to my example application. I believe elsewhere this is known as a "prior knowledge" situation. I ranecho http://localhost:8080/example/data | vegeta attack -duration 2s -rate 1/2s -h2c | tee results.bin | vegeta report
following the examples, and (hopefully) tellingvegeta
to send only one request through (my-rate
option is intended to express "one request every two seconds", and two seconds is the duration of the attack). I expected that from looking at my application's verbose logs I would see the same request as issued bycurl --http2-prior-knowledge http://localhost:8080/example/get
.Actual Behaviour
My application's logs show a different request coming in. Here is the initial log of the
vegeta
request (I am using a Netty-based server; these logs are what Netty can output to help diagnose issues):Here are the
curl
logs:Steps to Reproduce
See above; you should be able to see differences with any arbitrary HTTP/2-aware server/endpoint.
Additional Context
(No additional Github issues.)
I raise this because
vegeta
reports that there was a timeout while reading the body:…whereas the
curl
request has no such issues.I am no kind of HTTP/2 expert so perhaps the problem lies elsewhere.
Thanks for a neat little tool.